tutorial1.md

Hypergan tutorial 1

Thanks for trying out hypergan. This tutorial will walk you through creating your first model, exploring it, and building/deploying the generator.

Acquiring a dataset

You can acquire a dataset by having a folder full of loose images.

TODO: instructions for downloading dataset

Getting started

First install tensorflow and hypergan.

Testing install

To test tensorflow is installed correctly, start a python session and run import tensorflow as tf.

To test hypergan is installed correctly, run hypergan new . -l

You can view video card usage with nvidia-smi

Training hypergan

You can run:

  hypergan train -c online256 -s 256x256x3 -format png --sampler static_batch --sample_every 10 /path/to/data

Once hypergan is working, you can sample your

Sampling your model

You've created an infinite manifold. Great! With any luck it's diverse. You can see how diverse your model is by using:

  hypergan sample -c online256 -s 256x256x3 -format png --sampler random_walk /path/to/data

Building your model

  hypergan build -c online256 -s 256x256x3 -format png --sampler batch --sample_every 10 /path/to/data

Deploying your model

You've made a great model you'd like to share. Awesome! You can deploy it to users through a few methods.

Deploy to android

TODO: instructions

Come on over to our discord and share what you've created!